Predicting soil erosion and sediment yield at the basin scale: Scale issues and semi-quantitative models

Basin sediment yield is the product of all sediment producing processes and sediment transport within a basin. Consequently, the prediction of basin sediment yield should take into consideration all different erosion and sediment transport processes. However, traditional physics-based, conceptual, and empirical or regression models have not been able to describe all these processes due to insufficient systems knowledge and unfeasible data requirements. Therefore, the applicability of these models at the basin scale is troublesome. This paper first illustrates the relation between basin area, dominant erosion processes, and sediment yield by a combination of measured sediment yield at different spatial scales in Mediterranean environments. This clearly reveals that soil erosion rates measured at one scale are not representative for sediment yield at another scale level. Second, the most important semi-quantitative models developed for erosion and sediment yield assessments at the basin scale are reviewed. Most of these models use environmental factors to characterise a drainage basin in terms of sensitivity to erosion and sediment transport. Six of the nine models discussed (PSIAC, FSM, VSD, Gavrilovic, CSSM, WSM) include sheet-, rill-, gully, bank erosion, landslides, and connectivity, at least partly, in the assessment of basin sediment yield. The low data requirements and the fact that practically all significant erosion processes are considered makes them especially suited for estimating off-site effects of soil erosion. The other three models (EHU, CORINE, FKSM) focus mainly on sheet and rill erosion and provide quantitative descriptions of the sensitivity to erosion at basin or even regional scales. These models thus focus mainly on on-site problems of soil erosion. Most of the semi-quantitative models might benefit from a more quantitative description of factors used to characterise the basin. Though an equilibrium should be found between the extra effort and increase in model performance, the increased availability of spatially distributed topographic data as well as high-resolution satellite imagery will probably make this feasible in the near future.